Does any data leave my machine?

No. If you use the on-device engines (Apple Intelligence or local LLMs), no data ever leaves your computer. If you choose to use DeepL or OpenAI API, data is sent to their servers for translation only.

Are API costs (DeepL/OpenAI) included in the subscription?

No. LocaleKit is a 'Bring Your Own Key' (BYOK) tool. We provide the automation platform, but if you choose to use 3rd-party engines like DeepL, OpenAI, or Claude, you will need to use your own API keys and pay those providers directly for their usage.

Which frameworks do you support?

We support native iOS (Swift/Obj-C), Android (Kotlin/Java), Flutter, React/Web (JSON, YAML, PO), and backend files. We are constantly adding more formats.

Yes! The MAX plan includes a CLI tool that runs in GitHub Actions, GitLab CI, Bitrise, and CircleCI to validate and translate strings automatically.

Do I need English as base language?

Not necessarily, but it is recommended. You can configure any language as your source of truth in the workspace settings.

Do you preserve ICU messages?

Absolutely. We have specialized parsers that detect ICU syntax ({count, plural, ...}) and protect it from being mangled by translation engines.

How is the translation quality?

European languages are typically very good. Asian languages are accurate but can be too formal. This is why we provide a way to BYOK (Bring Your Own Key) with remote LLMs or DeepL, ensuring you can always choose the best engine for your specific needs.

Built for Developers

CLI tools, CI/CD pipelines, and deep technical integration for your localization workflow.

How it works internally

LocaleKit treats your code with the same care you do. It parses, protects, and reconstructs files to ensure zero syntax errors.

Developer Focused

We built this because we were tired of breaking builds with bad translations.

Scan & Detect

Recursively scans your workspace folder to identify localization file structures.

Parse Entries

Extracts keys, values, and comments into a unified internal format.

Protect Placeholders

Identifies and locks variables, ICU formatting, and HTML tags from translation.

AI Translation

Sends protected strings to the local LLM or your own API keys (DeepL, OpenAI, Claude).

Rebuild Files

Reconstructs original file formats with preserved indentation and comments.

Normalize for Git

Sorts keys deterministically to ensure clean diffs in your version control.

Automate with the CLI

Integrate LocaleKit into your workflow. One command to translate, validate, and sync your entire codebase.

Terminal — localekit

Translate missing keys to any language with your preferred engine.

$localekit translate --languages fr,de,ja --mlx-model mlx-community/Qwen3-4B-4bit

Loading MLX model mlx-community/Qwen3-4B-4bit...

✓ Model loaded (on-device, no data leaves your Mac)

Scanning workspace...

Found 3 localization files (.xcstrings)

Translating 142 keys → French...

Translating 142 keys → German...

Translating 142 keys → Japanese...

✓ 426 translations completed

✓ All placeholders preserved

✓ Files written successfully

Works with GitHub Actions, GitLab CI, Bitrise, and CircleCI.GitHub ActionsGitLab CIBitriseCircleCI

Choose your translation engine

Run models locally on your Mac with MLX — or bring your own API key for cloud engines. Your data, your choice.

No API key. No cloud. Everything stays on your Mac.

MLX

Local LLMFree

Run open-source models directly on Apple Silicon via the MLX framework. Supports Qwen3, Mistral, and Gemma models from Hugging Face.

Apple Intelligence

On-DeviceFree

Use Apple's built-in Foundation Models for context-aware translations. Requires macOS 15.1+ with Apple Intelligence enabled.

Models imported directly from Hugging Face:

26 models

TranslateGemma-4B·4B·3 GB

TranslateGemma-12B·12B·7 GBREC

TranslateGemma-27B·27B·16 GB

EuroLLM-22B (4-bit)·22B·13 GB

EuroLLM-22B (8-bit)·22B·24 GB

Qwen3-0.6B·0.6B·1 GB

Qwen3-1.7B·1.7B·2 GB

Qwen3-4B·4B·8 GB

Qwen3-8B·8B·16 GB

Qwen3-30B-A3B·30B MoE·16 GB

$0No API costs. Runs entirely on your Mac.

Use your own API keys. We never store or proxy your requests.

OpenAI

GPTBYOK

High-quality translations powered by GPT models. Excellent for context-aware, natural-sounding output across all languages.

DeepL

Neural MTBYOK

Industry-leading accuracy for European languages. Built-in glossary support for consistent terminology across projects.

BYOK = Bring Your Own Key. Use your own API keys from OpenAI or DeepL. No markup, no proxy.